Search

In conventional multi-user multiple-input multiple-output (MU-MIMO) systems with frequency division duplexing (FDD), channel acquisition and precoder optimization processes have been designed separately although they are highly coupled. This paper studies an end-to-end design of downlink MU-MIMO systems which include pilot sequences, limited feedback, and precoding. To address this problem, we propose a novel deep learning (DL) framework which jointly optimizes the feedback information generation at users and the precoder design at a base station (BS). Each procedure in the MU-MIMO systems is replaced by intelligently designed multiple deep neural networks (DNN) units. At the BS, a neural network generates pilot sequences and helps the users obtain accurate channel state information. At each user, the channel feedback operation is carried out in a distributed manner by an individual user DNN. Then, another BS DNN collects feedback information from the users and determines the MIMO precoding matrices. A joint training algorithm is proposed to optimize all DNN units in an end-to-end manner. In addition, a training strategy which can avoid retraining for different network sizes for a scalable design is proposed. Numerical results demonstrate the effectiveness of the proposed DL framework compared to classical optimization techniques and other conventional DNN schemes. [ABSTRACT FROM AUTHOR]

Channel covariance matrix (CCM) is one critical parameter for designing the communications systems. In this paper, a novel framework of the deep learning (DL) based CCM estimation is proposed that exploits the perception of the transmission environment without any channel sample or the pilot signals. Specifically, as CCM is affected by the user’s movement, we design a deep neural network (DNN) to predict CCM from user location and user speed, and the corresponding estimation method is named as ULCCME. A location denoising method is further developed to reduce the positioning error and improve the robustness of ULCCME. For cases when user location information is not available, we propose an interesting way that uses the environmental 3D images to predict the CCM, and the corresponding estimation method is named as SICCME. Simulation results show that both the proposed methods are effective and will benefit the subsequent channel estimation. [ABSTRACT FROM AUTHOR]

In this paper, we propose a deep learning-based beam tracking method for millimeter-wave (mmWave) communications. Beam tracking is employed for transmitting the known symbols using the sounding beams and tracking time-varying channels to maintain a reliable communication link. When the pose of a user equipment (UE) device varies rapidly, the mmWave channels also tend to vary fast, which hinders seamless communication. Thus, models that can capture temporal behavior of mmWave channels caused by the motion of the device are required, to cope with this problem. Accordingly, we employ a deep neural network to analyze the temporal structure and patterns underlying in the time-varying channels and the signals acquired by inertial sensors. We propose a model based on long short term memory (LSTM) that predicts the distribution of the future channel behavior based on a sequence of input signals available at the UE. This channel distribution is used to 1) control the sounding beams adaptively for the future channel state and 2) update the channel estimate through the measurement update step under a sequential Bayesian estimation framework. Our experimental results demonstrate that the proposed method achieves a significant performance gain over the conventional beam tracking methods under various mobility scenarios. [ABSTRACT FROM AUTHOR]

In this paper, we consider massive multiple-input-multiple-output (MIMO) communication systems with a uniform planar array (UPA) at the base station (BS) and investigate the downlink precoder design with imperfect channel state information (CSI). By exploiting channel estimates and statistical parameters of channel estimation error, we aim to design precoding vectors to maximize the utility function on the ergodic rates of users subject to a total transmit power constraint. By employing an upper bound of the ergodic rate, we leverage the corresponding Lagrangian formulation and identify the structural characteristics of the optimal precoder as the solution to a generalized eigenvalue problem. The Lagrange multipliers play a crucial role in determining both precoding directions and power parameters, yet are challenging to be solved directly. To figure out the Lagrange multipliers, we develop a general framework underpinned by a properly designed neural network that learns directly from CSI. To further relieve the computational burden, we obtain a low-complexity framework by decomposing the original problem into computationally efficient subproblems with instantaneous and statistical CSI handled separately. With the offline pre-trained neural network, the online computational complexity of precoder is substantially reduced compared with the existing iterative algorithm while maintaining nearly the same performance. [ABSTRACT FROM AUTHOR]

Reconfigurable intelligent surface (RIS) is a revolutionary technology for achieving high rate and large coverage in future wireless networks by smartly reflecting the signals with adjustable phase shifts. To design the reflection beamforming, accurate individual channel state information is required at the RIS, which is a challenge task due to the lack of signal processing ability in passive mode. In this paper, we add signal processing units for a few antennas at the RIS to partially acquire the channels and extrapolate them to the full channels, in which the active antenna selection is a key point but has not been addressed yet. We construct an active antenna selection network that utilizes the probabilistic sampling theory to select the optimal locations of these active antennas. With this active antenna selection network, we further design two deep learning-based schemes, i.e., the channel extrapolation scheme and the beam searching scheme. The former utilizes the selection network and a convolutional neural network to extrapolate the full channels from the partial channels, while the latter adopts a fully-connected neural network to achieve the direct mapping from the partial channels to the optimal beamforming vector with maximal transmission rate. Simulation results show that the proposed optimal antenna selection outperforms the trivial uniform antenna selection, and the performance of beam searching is more stable than that of channel extrapolation with fewer active antennas. [ABSTRACT FROM AUTHOR]

In frequency-division duplexing systems, the downlink channel state information (CSI) acquisition scheme leads to high training and feedback overhead. In this work, we propose an uplink-aided downlink channel acquisition framework using deep learning to reduce such overhead. We consider the entire downlink CSI acquisition process, including the downlink pilot design, channel estimation, and feedback. First, we propose an adaptive pilot design module by exploiting the correlation in magnitude among bidirectional channels in the angular domain to improve channel estimation. Second, to avoid the bit allocation problem during the feedback module, we concatenate the complex channel and embed the uplink channel magnitude to the channel reconstruction at the base station. Finally, we combine the two modules and compare two popular uplink-aided downlink channel acquisition frameworks. One framework estimates and subsequently feeds back the channel at the user equipment. In the other framework, the user equipment directly feeds back the received pilot signals to the base station. Results reveal that with the help of the uplink channel, directly feeding back pilot signals can save approximately 20% of feedback bits. This work thus provides a guideline for future research. [ABSTRACT FROM AUTHOR]

In order to reduce the downlink training overhead of mmWave communications, we propose a novel downlink beamforming strategy using the uplink sub-6GHz channel and downlink mmWave pilots that are sent from a few active antennas. Specifically, we design a novel dual-input neural network architecture, called FusionNet, to merge the sub-6GHz channel and the channel of a few active mmWave antennas. The proposed fusion model could intelligently adjust the attention paid (by the neural network) for sub-6GHz channel and mmWave channel by an attention mechanism. The output of the FusionNet represents the probability for each beam being the optimal one. We also propose an antenna selection model that can choose better active antennas to send the downlink pilots, in which the gradient of antenna selection vector is approximated by that of an antenna probability vector. Simulation results demonstrate the superior performance of the proposed strategy compared to the existing one that purely relies on the sub-6GHz information or compared to the shallow model that directly adds uniform pilots. [ABSTRACT FROM AUTHOR]